An Industrial-Scale Sequential Recommender for LinkedIn Feed Ranking

TL;DR

This paper introduces Feed-SR, a transformer-based sequential recommender system for LinkedIn Feed, which improves user engagement and meets strict production constraints through optimized modeling, training, and deployment techniques.

Contribution

The paper presents Feed-SR, a novel transformer-based sequential ranking model optimized for large-scale deployment in LinkedIn Feed, outperforming previous models in engagement and efficiency.

Findings

Feed-SR increases member engagement by +2.10% in online tests.

Feed-SR outperforms alternative architectures in online metrics and efficiency.

Successful deployment of a large-scale transformer-based recommender system.

Abstract

LinkedIn Feed enables professionals worldwide to discover relevant content, build connections, and share knowledge at scale. We present Feed Sequential Recommender (Feed-SR), a transformer-based sequential ranking model for LinkedIn Feed that replaces a DCNv2-based ranker and meets strict production constraints. We detail the modeling choices, training techniques, and serving optimizations that enable deployment at LinkedIn scale. Feed-SR is currently the primary member experience on LinkedIn's Feed and shows significant improvements in member engagement (+2.10% time spent) in online A/B tests compared to the existing production model. We also describe our deployment experience with alternative sequential and LLM-based ranking architectures and why Feed-SR provided the best combination of online metrics and production efficiency.